Tomasz Tunguz’s Post

Select avg(Moby Dick) limit 2 sentences The SQL statement above is a quote from our recent Office Hours with Benn Stancil. It’s not a SQL statement that would work today in a cloud data warehouse. But an LLM would understand it : summarize the book Moby Dick in two sentences. Sure enough, ChatGPT answers the question : This pseudocode blends the structured queries of data analysis with the unstructured data contained in a classic novel. This is how Benn views the future of BI BI’s Third Form. “Is the output numbers in the future of BI? I don’t think it’s numbers…the theory is what we want ; the characterization & qualitative parts.” Just as ChatGPT summarized the essence of Moby Dick, AI easily summarizes the themes of 10 customer conversations - or 100. Benn views the future of BI as the narrative, the story behind the data. Numbers have been the historical source of truth, a proxy to underlying customer behavior. Nobel laureate Shiller wrote about the power of narrative economics, stories used to understand a business & shape the future. The rest of the Modern Data Stack will transform, too. Data Quality will need to tackle ever greater volumes. If LLMs do average the insight in new underlying unstructured data sources which are potentially massive, what is Data Quality’s new role in the world of AI? ETL, more deterministic in nature, may not see full delegation to AI. We covered the Databricks/Tabular acquisition & many other topics. Benn’s views on the data ecosystem are unique : a bit irreverent with lots of practical experience underpinning them. I hope you enjoy the conversation as much as I did. The video link is here : https://lnkd.in/gVPK4X2K and the podcast link is here : https://lnkd.in/gv5R5xrZ.

Select avg(Moby Dick) limit 2 sentences The SQL statement above is a quote from our recent Office Hours with Benn Stancil. It’s not a SQL statement that would work today in a cloud data warehouse. But an LLM would understand it : summarize the book Moby Dick in two sentences.

“This is the first of your newsletters that doesn’t align well with what I’ve been seeing in the field.” After publishing The Four Barriers to AI Adoption, David Morse, a reader & a friend who was most recently CRO at Hebbia & VP Sales at Scale AI sent me this email. Dave continued : “The biggest blocker to adoption at AI application companies is user education and limitations of frontier models. Finding use cases that work; steering users away from failure cases. Prompting for use cases that work. Dealing with stochasticity. This is where agents comprise most good adoption stories I know of. GPTs get into production quickly while search based apps or chatbots have a longer path to wide adoption. Even small startups can push agents into production in <2 weeks. Most customers will waive or streamline security review, procurement processes, etc. to get their desired solution in house IF the vendor can demo a use case that actually works. They will gate some data access until the vendor has certifications or the business simply pushes IT to approve access. Some companies are closing deals in <90 days with no demo…there is some impressive selling going on. Most AI companies I know send their initial MSA waiving rights to train on customer data. This can be added later if the business needs it. With issues around data retention and training waived the MSA doesn’t look that different from data analytics companies. Customer legal teams push risk around data leaks or other data issues into higher limit of liability” One of the gifts of writing in public is receiving emails like this one that offer a different view of what’s working in the field. Some things stand out to me from Dave’s perspective : The nuance in selling between GPTs compared to search & chatbots is a great example of how the market is evolving differently in different segments : lumping all AI solutions into the same category is a mistake. Leveraging the executive sponsorship to bypass procurement processes is consistent with the AI imperative boards & executives have championed. The AI market isn’t a single market : it’s reflective of the software market both older segments & newer segments. If you have a view on how the AI software market is evolving, I’d love to hear from you.

“This is the first of your newsletters that doesn’t align well with what I’ve been seeing in the field.” After publishing The Four Barriers to AI Adoption, David Morse, a reader & a friend who was most recently CRO at Hebbia & VP Sales at ScaleAI sent me this email.

AI adoption is slower than expected in many spaces. Some of the reasons are straightforward, but others are more subtle. Most leaders wants to inject AI into their business to develop a competitive advantage. There are four challenges. 1. The first challenge is understanding the technology’s ability. Because the capabilities evolve so quickly, it’s hard to keep up. If PhDs in the domain are rushing to understand the capabilities reading papers every week, how are business leaders meant to grok the state of the art? Also, because the systems are non-deterministic, they are unpredictable. The pace of innovation, the early understanding of AI internals, & the non-determinism compound to create doubt. 2. Security is the second challenge. Because of their unpredictable nature & because few have expertise launching these systems, product managers, engineering leaders, & security teams are hesitant to launch both internal & external systems until they develop confidence in data security. AI security has at least four dimensions : model security, prompt injection, RAG authentication/authorization, & data loss prevention. 3. Legal is the next barrier to entry. Master service agreements (MSAs) are the contracts that dictate terms of service, data privacy, & service levels between a buyer & a vendor. These agreements’ clauses are well-trodden & known. AI is new. Should a company allow a vendor to train a model using their data? Whose intellectual property is a fine-tuned model? What happens if a vendor violates the data privacy law? What training data is used that might subject the software buyer to future legal action? Many legal teams are working to understand those questions. 4. Procurement is yet another barrier. SOC2, GDPR, ISO27001 & other certifications provide industry standards for security & compliance. But no such standard exists for AI - yet. Bias, fairness, & explainability are all important factors in AI : some are important for public relations, others for compliance. Selling AI is not just selling software. Many of the processes are new & these barriers introduce friction into the sales process, extending sales cycles. Over time, these rough edges will be worn smooth through practice. But the first companies selling today will need to persist through these challenges.

AI adoption is slower than expected in many spaces. Some of the reasons are straightforward, but others are more subtle. Most leaders wants to inject AI into their business to develop a competitive advantage. There are four challenges.

It’s time for the 2024 Annual Theory Go-to-Market Survey. This is a brief 28-question survey : https://lnkd.in/ggBJv7XN. Our goal is to understand how startups have evolved their sales, marketing, customer success, and cash management over the last four years by comparing these results to those through the go-go years of 2020 and beyond. We will publish these results and answer questions about them at upcoming Office Hours. With this data, we should be able to draw some broader conclusions about the shift from growth to efficiency & determine if the buyer behavior changes in the private market parallel those in the public market. In addition, we should be able to draw some initial inferences about the impact of AI on sales teams. If you complete the survey, I will share with you the anonymized raw data so you can perform your own analyses. If you have questions, just message me on Twitter or send me an email.

It’s time for the 2024 Annual Theory Go-to-Market Survey. This is a brief 28-question survey. Our goal is to understand how startups have evolved their sales, marketing, customer success, and cash management over the last four years by comparing these results to those through the go-go years of 2020 and beyond. We will publish these results and answer questions about them at upcoming Office Hours.

What does it take to go public? Has it changed over the last 15 years? We gathered data on the US venture-backed software companies that went public between 2010 & today. We corrected the trailing 12 months’ revenue at the time of IPO for inflation & plotted the data. Before 2018, only one company IPOed with more than $200m in revenue. In fact, the median revenue at IPO at $90m. Today, the median revenue at IPO is $189m (corrected for inflation), more than double. This ratio is roughly true across the board for 25th & 75th percentile-sized companies as well. If we divide the data set into two eras : before 2018 & 2018 onwards, we see a clear shift in the average revenue at IPO & the difference is statistically significant with a p value of 0.0063. Before 2018, 11 companies went public with less than $100m in trailing revenue. Afterwards, only 2. Why such a difference? IPOs are extremely expensive. A company with about $100m in revenue will pay between $9-26m to go public. For a company generating $100m a year, this means an IPO cost 16% of their revenue to raise a round of capital : likely the most expensive capital raise of the company’s life. For a later-stage company, the costs of a private financing might may touch $1m in the upper reaches, but is much more likely to be measured in the hundreds of thousands of dollars. As private dollars entered the late stage market, why wouldn’t a startup raise private dollars instead of public? That’s precisely the pattern the data illustrates. Private market capital values companies at higher prices (from higher multiples) & is less expensive to raise. While there are still many companies sub-$200m in revenue who go public, the late-stage private markets enable those who would wait to achieve much higher scale. In addition, this data doesn’t account for the many large companies who have yet to IPO but very likely will : Databricks, Stripe, Canva, Rippling, OpenAI, etc., which will further underscore the trend. PwC estimates cost based on historical data from 2019-2022

What does it take to go public? Has it changed over the last 15 years?